10
Clinical Focus
Mandatory contrast-enhanced venography to detect deep-vein thrombosis (DVT) in studies of DVT prophylaxis: upsides and downsides Jack Hirsh; Jeffrey S. Ginsberg; Noel Chan; Gordon Guyatt; John W. Eikelboom Department of Medicine, McMaster University, Hamilton, Ontario, Canada
Summary The introduction of venography into patient care was a major advance because it was the first accurate method for the diagnosis of DVT. Compression ultrasound has since become the preferred test for patients with suspected DVT because, unlike venography, it is simple, non-invasive and widely available. Venography has facilitated the development and approval of new anticoagulants and remains widely used as an efficacy outcome in trials of venous thromboembolism prevention. Most thrombi detected by screening venography are, however, small and unimportant for patients. In order to calculate the
Correspondence to: John W. Eikelboom Department of Medicine, McMaster University Hamilton General Hospital 237 Barton Street East Hamilton, Ontario, L8L 2X2, Canada E-mail:
[email protected]
Venography as a diagnostic test for suspected (symptomatic) venous thrombosis Less than 50 years ago, patients with suspected deep-vein thrombosis (DVT) were diagnosed on clinical grounds without the aid of objective diagnostic testing. Reports that many patients with fatal venous thromboembolism (VTE) were not diagnosed ante-mortem (1-3) and that only about one-half of patients with a clinical suspicion of DVT had DVT (4) challenged the accuracy of clinical diagnosis. The uptake of venography to diagnose DVT was slow, both because its use challenged a time-honoured clinical approach, and because the test was not widely available. Clinicians accepted the presence of a persistent intraluminal filling defect on contrast venography as being diagnostic of DVT, and therefore the need to treat such patients with anticoagulants to prevent thrombus progression and reduce the risk of complications (pulmonary embolism [PE] and the post-thrombotic syndrome). Clinicians questioned, however, the validity of a normal venogram, particularly when this result contradicted their clinical judgment. They were ultimately convinced when a study demonstrated that it is safe to withhold anticoagulants in patients with suspected venous thrombosis who had a normal venogram (at 3 month follow up, 2 out of 160 patients with negative venogram suffered from a DVT [odds ratio [OR] 1.3%; 95% confidence interval [CI]: 0.15% to 4.4%]) (5). Thus, by introducing venography into routine clinical practice, © Schattauer 2014
trade-off between an asymptomatic thrombus and a bleed we require an estimate of the number of asymptomatic thrombi that must be prevented to avoid a patient-important thrombus. A credible estimate of this ratio is not available. Therefore when used as a measure of efficacy in trials of thromboprophylaxis, venography has limitations for comparing the relative effects of alternative antithrombotic agents on outcomes important to patients.
Keywords Deep-vein thrombosis, venous thrombosis, prophylaxis
Received: July 12, 2013 Accepted after major revision: August 30, 2013 Prepublished online: October 10, 2013 doi:10.1160/TH13-07-0562 Thromb Haemost 2014; 111: 10–13
many patients suspected of having DVT were spared unnecessary treatment with anticoagulants. Venography is not without its problems. Apart from being invasive, expensive, and sometimes painful, in 10-20% of patients its interpretation is impossible for technical reasons (e.g. inability to cannulate a vein, deep veins of the leg are not filled adequately) (6). In virtually all diagnostic imaging departments, venography has been replaced by venous compression ultrasound (CUS) in patients with symptoms suspicious for DVT. CUS does not always reliably detect distal (calf vein) DVT and some small non-occlusive proximal (popliteal and/or more proximal vein) thrombi that are routinely identified by venography (7). This limitation is not problematic because if initial CUS is normal and remains normal when repeated in about 7 days (8, 9), or if an initial whole leg CUS is negative (10), it is safe to withhold treatment. At the same time, however, these findings highlight that the vast majority of calf vein and small non-occlusive proximal vein thrombi, which comprise a large proportion of post-operative venous thrombi, are unimportant for patients (11).
Venography as an outcome measure in clinical trials for the prevention of VTE The most important remaining role of venography, and the main subject of this review, is to improve the efficiency of clinical trials Thrombosis and Haemostasis 111.1/2014
11
Hirsh et al. Mandatory contrast-enhanced venography to detect DVT in studies of DVT prophylaxis
evaluating prophylactic agents in patients at high risk of VTE. Over the years, several tests have been used to screen high risk patients for DVT, including radio-labelled fibrinogen uptake scanning (RFUS), impedance plethysmography (IPG), CUS and venography. Although venography is the most invasive and least convenient, it has prevailed as the method of choice because it proved to be more specific than RFUS (particularly after hip surgery) and more sensitive than IPG and CUS. Most thrombi detected by screening venography are small, distal, and asymptomatic, but a small minority is large and can be complicated by major and even fatal PE. Routine screening venography to detect and treat DVT before it causes fatal PE is inappropriate because it is probably less effective than primary prophylaxis, and certainly not cost-effective. On the other hand, screening venography has provided valuable information about the epidemiology and distribution of DVT in high-risk medical and surgical patients, and has played important roles in informing dose selection and the clinical approval process for new anticoagulants. Phase 2 dose-finding studies would be much more challenging without the use of screening venography and phase 3 studies that led to the approval of low molecular weight heparins, fondaparinux and the new oral anticoagulants (rivaroxaban, dabigatran, and apixaban) would have been prohibitively expensive without availability of screening venography to assess efficacy outcome. Despite its importance in anticoagulant development, screening venography has major limitations when used as a surrogate for patient-important VTE (12, 13). This limitation of screening venography is illustrated by the following hypothetical example of choosing among prophylactic agents evaluated in phase 3 trials. If one antithrombotic agent "A" is more effective than another agent "B" for reducing asymptomatic venous thrombosis, but A causes more bleeding it would be inappropriate to conclude that "A" is preferred over "B". Such a recommendation could only be made if: 1) "A" caused no more bleeding than "B" and was associated with a reduced risk of objectively diagnosed symptomatic DVT; or 2) the relative patient-importance (disutility) of an asymptomatic venographically-detected thrombosis and a bleeding episode were known. Support for this viewpoint comes from the impact of the clinical trials comparing the then established anticoagulant enoxaparin, with the then new anticoagulant, fondaparinux, in patients undergoing major hip or knee surgery. The program comprised four randomized trials and included 7,344 patients undergoing major hip or knee surgery (14). The primary efficacy outcome was total VTE up to day 11, defined as DVT detected by mandatory venography plus symptomatic DVT and PE, and the main safety outcome was major bleeding. Meta-analysis of the four trials showed that total VTE rates were more than halved (6.5% vs 13.5%, p
